Time Usage

[blazespinnaker]

I was wondering if there was a good breakdown of the time-usage stats anywhere. I see that entity update is quite high, assuming that's assemblers and such. But electric network surprised me a bit. Was wondering what caused that to be high.

[blazespinnaker]

zowie: https://mulark.github.io/test-index.html

[blazespinnaker]

I am a bit surprised at how much electricity consumes in terms of the game loop. it seems like the sort of thing you could check every 1-2 seconds per entity without any noticeable change in game play. Calculating consumption per entity should just be an issue of incrementing counters. Also, sadly, when you power entities off they continue to consume CPU time.

[Koub]

Don't forget production counts too. And also, if there are power switches that shut down part of the factory, THAT is a resource hog.
To control production it's advised to block feeding belts instead of unpowering whole parts of the factory (CPU-wise).

[blazespinnaker]

Yeah, I was looking more at using substations and radar. And "underpowering" is not quite right, but rather zero power / turning off. For example, I could have a power switch that turns off radar on 1/4 of the map and I just turn it on when I want to look at it.

Unfortunately, after turning off the power I didn't see a noticeable improvement in update lag.

i guess it's just not something people do, so it was never optimized for.

[Koub]

In the past, the devs have explained that unpowering entities that may go to sleep (typically assembling machines, inserters, ...) keeps consuming CPU, while they stop doing so when they go to sleep mode because of input starvation or output saturation.

[Rseding91]

High electric network is almost always because someone has built a boat-load of isolated networks instead of one-big-network. For example: 10~ solar panels and a radar, copy-pasted around. There's a fixed overhead for an electric network to exist and the per-entity cost gets cheaper the more entities are in a given network.

[starlinvf]

High electric network is almost always because someone has built a boat-load of isolated networks instead of one-big-network. For example: 10~ solar panels and a radar, copy-pasted around. There's a fixed overhead for an electric network to exist and the per-entity cost gets cheaper the more entities are in a given network.

So whats the overhead cost for enable/disable on entities? Cuz I'm only just realizing fluid handling objects, like pump, can be hooked up to a circuit network. Up until now, I've been controlling power to clusters of Chem plants.... but never thought of a potential performance impact.

[Impatient]

High electric network is almost always because someone has built a boat-load of isolated networks instead of one-big-network. For example: 10~ solar panels and a radar, copy-pasted around. There's a fixed overhead for an electric network to exist and the per-entity cost gets cheaper the more entities are in a given network.

Oh my, self sustaining radar outposts are exactly what I am using in all my games. That's good to know.

[eradicator]

High electric network is almost always because someone has built a boat-load of isolated networks instead of one-big-network. For example: 10~ solar panels and a radar, copy-pasted around. There's a fixed overhead for an electric network to exist and the per-entity cost gets cheaper the more entities are in a given network.

Would a stand-alone void energy source radar prevent that problem?

[starlinvf]

High electric network is almost always because someone has built a boat-load of isolated networks instead of one-big-network. For example: 10~ solar panels and a radar, copy-pasted around. There's a fixed overhead for an electric network to exist and the per-entity cost gets cheaper the more entities are in a given network.

Would a stand-alone void energy source radar prevent that problem?

Or a programmable power source you can designate being connected to an electrical network, but remotely?